Bug? in os.chown()

[Jim Dennis]

In article <slrn95195m.hn4.qrczak at qrnik.zagroda>, 
 Marcin 'Qrczak' Kowalczyk wrote:

>01 Jan 2001 15:09:07 +0100, Martin von Loewis <loewis at informatik.hu-berlin.de> pisze:

>> To    change    the    symbolic    link    itself,    use
>> lchown(2).

>It should not be necessary. Symbolic link mode is not relevant anyway.

 It should not *normally* be necessary.  However, there are
 some special cases under Linux where the symlink permissions
 might be important.  In particular there is a set of patches
 from Solar Designer (don't know his real name) which allow one
 to impose a set of additional (non-POSIX) semantics on symlinks.
 (These prevent most symlink attacks from working, since it
 causes the kernel to refrain from following symlinks from 
 sticky directories (such as /tmp) unless that symlink is owned
 by root or the symlink and its target have matching owners).

 I'm sure there are other, probably more common cases where the 
 link ownership might be important.  I don't know of any cases
 under Linux where link *permissions* are considered.